Apple Tester Skill

SkillAI & models

CLI-first Apple UI tester for iOS Simulator and macOS app automation. Use for deterministic observe/act/verify testing and promotion to XCTest artifacts.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Apple Tester Skill skill

What this skill tells your AI

The instructions your AI receives, as published by lev-os/agents in skills-db/apple/apple-tester/SKILL.md and read by ahel’s review.

Purpose

apple-tester is a tester harness, not an autonomous runtime. Default usage is direct command execution + dev watch loop. Use server mode only when you need persistent session state or streamed logs.

Use apple-tester as the only public entrypoint. Do not invoke apple-driver (or agent-apple) directly for daemon/session workflows.

Strict loop

Always run this loop:

  1. snapshot
  2. Plan one action
  3. Execute one action
  4. Verify with is / get / assert
  5. Repeat

Never run free-form multi-action plans without verification.

Quick start

If apple-tester is not on your PATH yet, install it once from the repo root:

pnpm apple-tester:cli:install
apple-tester doctor --build-driver
apple-tester server start
apple-tester session start --macos
apple-tester snapshot --json
apple-tester tap @m1
apple-tester is visible @m1
apple-tester session stop
apple-tester server stop

Host input safety

macOS host-input gestures are blocked by default to avoid stealing local mouse/keyboard focus.

Use --allow-host-input only for explicit gesture actions:

apple-tester session start --macos --allow-host-input
apple-tester hold @m1 --duration-ms 900 --allow-host-input
apple-tester right-click @m3 --allow-host-input

iOS flow

apple-tester server start
apple-tester session start --ios --udid <SIM_UDID>
apple-tester snapshot --json
apple-tester tap @i1
apple-tester swipe --from 200,700 --to 200,250
apple-tester session stop
apple-tester server stop

macOS gestures

apple-tester hold @m1 --duration-ms 900 --allow-host-input
apple-tester longpress --coords 640,380 --allow-host-input
apple-tester right-click @m3 --allow-host-input
apple-tester swipe --from 400,500 --to 900,500 --duration-ms 500 --allow-host-input

Promotion

Promotion compiles a successful exploratory run into deterministic XCTest files and a promotion manifest:

apple-tester promote --run-id <RUN_ID> --platform ios --name "Login Smoke"
apple-tester promote --run-id <RUN_ID> --platform ios --name "Login Smoke" --dry-run --diff-lines 120
apple-tester promote --run-id <RUN_ID> --platform ios --name "Login Smoke" --dry-run --write-report ./artifacts/promotion-report.json

Output targets:

  • iOS: apps/ios/UITests/Generated/*.swift
  • macOS: apps/macos/UITests/Generated/*.swift

Log streaming for agents

For parser-friendly polling, prefer NDJSON log output:

apple-tester logs --json --follow | jq -c '.entries[] | {seq, level, message}'

Canonical local development loop

pnpm dev:watch

This performs deterministic change detection and targeted rebuilds without relaunching apps by default (--no-relaunch).

Notes

  • right-click is macOS-only.
  • hold/longpress use press-and-hold semantics on macOS and best-effort equivalent on iOS.
  • Prefer accessibility identifiers in app code for stable promoted tests.
  • Use apple-tester logs --follow for live daemon diagnostics when actions fail.

Technique Map

  • Identify scope — Determine what the skill applies to before executing.
  • Follow workflow — Use documented steps; avoid ad-hoc shortcuts.
  • Verify outputs — Check results match expected contract.
  • Handle errors — Graceful degradation when dependencies missing.
  • Reference docs — Load references/ when detail needed.
  • Preserve state — Don't overwrite user config or artifacts.

Technique Notes

Skill-specific technique rationale. Apply patterns from the skill body. Progressive disclosure: metadata first, body on trigger, references on demand.

Prompt Architect Overlay

Role Definition: Specialist for apple-tester domain. Executes workflows, produces artifacts, routes to related skills when needed.

Input Contract: Context, optional config, artifacts from prior steps. Depends on skill.

Output Contract: Artifacts, status, next-step recommendations. Format per skill.

Edge Cases & Fallbacks: Missing context—ask or infer from workspace. Dependency missing—degrade gracefully; note in output. Ambiguous request—clarify before proceeding.

Signals

GitHub stars
22
Forks
2
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
apple-tester
Source
github.com/lev-os/agents